[Identification and interfering factors of falsely elevated androgen levels in clinical laboratory testing in women]

Objective: To assess discrepancies in androgen measurements, investigate potential interfering factors, and propose an optimized diagnostic pathway for women with falsely elevated androgen levels. Methods: Thirty female cases with falsely elevated androgen levels at Peking University First Hospital between October 2021 and September 2025 were included to compare chemiluminescence immunoassay (CLIA) and liquid chromatography-tandem mass spectrometry (LC-MS/MS) in measuring testosterone (T), androstenedione (A4), and dehydroepiandrosterone sulfate (DHEAS). We assessed methodological agreement using Passing-Bablok regression and Bland-Altman analysis. Heterophilic antibody (HA) blocking and liquid-liquid extraction were performed to identify potential interfering factors. Results: The mean patient age was (31.2±6.3) years. Cutaneous hyperandrogenic signs were absent in 63.3% (19/30) of patients, ovarian imaging abnormalities were absent in 60.0% (18/30), and adrenal imaging performed in 43.3% (13/30) of cases revealed no abnormalities. Total testosterone levels measured by the Beckman, Siemens, and Roche CLIA platforms were 1.12, 0.32, and 0.46 ng/ml, respectively, each significantly exceeding the LC-MS/MS value of 0.27 ng/ml (all P<0.05). Moreover, A4 and DHEAS levels determined by the Siemens platform were 2.21 ng/ml and 2 660 μg/L, respectively, also significantly higher than the corresponding LC-MS/MS results of 1.07 ng/ml and 2 164 μg/L (both P<0.05). The rate of false elevation in T varied significantly across different CLIA platforms. The false elevation rate for A4 was 16.7%(5/30), while no false elevation was observed for DHEAS. CLIA assays showed systematic positive biases for all three androgens tested, with median relative biases ranging from 26.0% to 352.0%, 69.0%, and 19.0%, respectively. The proportion of patients exceeding the allowable total error (TEa,±15%) was 76.2% to 100.0%, 93.3%, and 56.6% for the three analytes, respectively. Passing-Bablok regression indicated a significant constant positive bias for T and A4, as measured by CLIA, with proportional bias also observed for A4. In contrast, no significant systematic or proportional bias was found for DHEAS. Bland-Altman analysis confirmed that the limits of agreement for all three androgens by CLIA exceeded the TEa range. Interference verification suggests that factors such as HA may be associated with false elevation of T. In cases of falsely elevated T levels, the low-dose dexamethasone suppression test effectively suppressed DHEAS, but the suppression rate of T lacked diagnostic value for differentiation. Conclusion: CLIA generally overestimates serum T and A4 levels in women than LC-MS/MS.